Anthropic reveals Claude Mythos Preview, withholds it over cyber risk and launches Project Glasswing
Confirmed
Importance: historic (5 of 5)The takeaway
On April 7, 2026 Anthropic disclosed Claude Mythos Preview, a general-purpose frontier model so strong at finding and exploiting software vulnerabilities that Anthropic declined to release it generally.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- Historic (5 of 5)
- Last verified
- 1 October 2026
Your AI and this story
- GPT-6 AstraIn its training data
- Claude Opus 5.5In its training data
- Gemini 4 Hydrogen7 days after its cutoff
- Grok 4.7In its training data
It is after the cutoff of Gemini 4 Hydrogen, and can be in the training data of GPT-6 Astra, Claude Opus 5.5 and Grok 4.7.
Key facts
- Announced April 7, 2026 after drafts leaked on March 26, 2026
- SWE-bench Verified 93.9% (Opus 4.6: 80.8%); SWE-bench Pro 77.8% (53.4%); Terminal-Bench 2.0 82.0% (65.4%); CyberGym 83.1% (66.6%)
- Exploit evals (Anthropic, May 22, 2026): arbitrary code execution on 21 of 41 V8 bugs in ExploitBench (other models at most 2); 157 intended-vuln successes on ExploitGym’s 898 tasks vs 15 for Opus 4.6; $35M of smart contracts exploited on SCONE-bench, ~75% more than the next model
- Found thousands of zero-days across major OSes and browsers: a 27-year-old OpenBSD remote-crash flaw, a 16-year-old FFmpeg bug, Linux kernel privilege escalations
- Glasswing launch partners: AWS, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks + 40 more
- $100M in Mythos Preview credits; $2.5M to Alpha-Omega/OpenSSF; $1.5M to Apache Software Foundation
- Participant pricing $25 / $125 per 1M tokens
- Mozilla later reported 271 Firefox vulnerabilities found with Mythos Preview (Apr 21); Glasswing grew from 50 to 200 organizations on June 2
What happened
Per Wikipedia’s timeline, the announcement set off a wave of government reactions. US Treasury Secretary Bessent and Fed Chair Powell convened financial executives on April 9. The White House met Anthropic on April 16. India’s Finance Ministry and Japan’s FSA held meetings on April 23–24, and 32 US Representatives wrote to the National Cyber Director on May 13. Wikipedia also reports that unauthorized users got access on launch day via details from the Mercor data breach.
Why it matters
Mythos Preview marked the point where a frontier lab judged a model’s offensive cyber capability too dangerous for general release. It shaped the rest of Anthropic’s 2026: the Fable/Mythos safeguard split, verification programs and export-control fights.
Sources
11 sources from 8 sites. Numbers match the chips in the text.
11 sources: 7 primary, 1 press, 3 reactions
Primary
- Project Glasswing (Anthropic)anthropic.com, official
- Assessing Claude Mythos Preview’s cybersecurity capabilitiesanthropic.com, official
- Anthropic: Measuring LLMs’ ability to develop exploits (May 22, 2026)anthropic.com, official
- Claude Mythos Preview’s cybersecurity capabilities (red.anthropic.com)red.anthropic.com, official
- Claude Mythos product pageanthropic.com, official
- AWS Bedrock model card: Claude Mythos Previewdocs.aws.amazon.com, docs
- Anthropic on X: Introducing Project Glasswing, powered by Claude Mythos Previewx.com, official
Press
- Google Cloud: Claude Mythos Preview on Agent Platformcloud.google.com, press
Reactions
- CETaS (Turing Institute): What does Mythos mean for cybersecurity?cetas.turing.ac.uk, discussion
- Wikipedia: Claude Mythosen.wikipedia.org, discussion
- Project Glasswing video (Anthropic)youtube.com, video
Changes
- Added May 22 exploit-evals post (ExploitBench/ExploitGym/SCONE-bench results)
- Added post link(s) (posts-as-events pass)
- Filed
